Method and apparatus for filtering compressed textures

Abstract

An apparatus and method are described for texture compression, decompression and filtering. For example, one embodiment of a method comprises: determining distances between each of a plurality of texels of a texture block and each of a plurality of approximation points; generating a decompression matrix comprising a plurality of radial basis function RBF values over the distances using a specified type of RBF; using the decompression matrix to generate a decompression-filtering matrix according to a defined filter function, the decompression-filtering matrix being usable to generate a decompressed and filtered version of the texture block as a result of the filter function being integrated into the decompression-filtering matrix.

What is claimed is: 
     
       1. A method comprising:
 determining distances between each of a plurality of texels of a texture block and each of a plurality of approximation points, wherein the approximation points comprise radial basis function (RBF) center points, and wherein each of the plurality of approximation points is selected from edges of the texture block; 
 generating a decompression matrix comprising a plurality of radial basis function (RBF) values over the distances; 
 generating a matrix of approximation coefficients using the decompression matrix and a set of texel color values sampled over the texture block, wherein the matrix of approximation coefficients represents a compressed version of the texture block; 
 using the decompression matrix to generate a decompression-filtering matrix according to a defined filter function, the decompression-filtering matrix being usable to generate a decompressed and filtered version of the texture block as a result of the filter function being integrated into the decompression-filtering matrix; and 
 performing texture mapping operations using the decompressed and filtered version of the texture block. 
 
     
     
       2. The method as in  claim 1  further comprising:
 generating the decompressed and filtered version of the texture block by performing a product of data from the decompression-filtering matrix and the matrix of approximation coefficients. 
 
     
     
       3. The method as in  claim 2  further comprising:
 reshaping the decompression-filtering matrix; and 
 performing the product of the reshaped decompression-filtering matrix and the matrix of approximation coefficients. 
 
     
     
       4. The method as in  claim 1  further comprising:
 determining a distance matrix [DM] using the distances between each of the plurality of texture block texels and each of the plurality of approximation points; and 
 determining the decompression matrix of member-wise RBF values over the distance matrix [DM] using a specified type of RBF and specified shape parameter e p  value. 
 
     
     
       5. The method as in  claim 4  wherein the specified type of RBF is selected from a group consisting of Gaussian and Multiquadratic radial basis functions. 
     
     
       6. The method as in  claim 1  wherein using the decompression matrix to generate the decompression-filtering matrix comprises performing linear combinations on sets of elements from the decompression matrix to arrive at new sets of elements for the decompression-filtering matrix based on the filter function. 
     
     
       7. The method as in  claim 6  wherein the linear combinations comprise taking an average value of multiple elements from the decompression matrix to determine each element in the decompression-filtering matrix. 
     
     
       8. The method as in  claim 6  wherein the operations cause the number of elements in the decompression-filtering matrix to be less than, greater than or equal to the number of elements in the decompression matrix. 
     
     
       9. A processor comprising:
 a first circuit to:
 determine distances between each of a plurality of texels of a texture block and each of a plurality of approximation points, wherein the approximation points comprise radial basis function (RBF) center points, and wherein each of the plurality of approximation points is selected from edges of the texture block; 
 generate a decompression matrix comprising a plurality of radial basis function (RBF) values over the distances using a specified type of RBF; 
 generate a matrix of approximation coefficients using the decompression matrix and a set of texel color values sampled over the texture block, wherein the matrix of approximation coefficients represents a compressed version of the texture block; and 
 use the decompression matrix to generate a decompression-filtering matrix according to a defined filter function, the decompression-filtering matrix being usable to generate a decompressed and filtered version of the texture block as a result of the filter function being integrated into the decompression-filtering matrix, 
 wherein the decompressed and filtered version of the texture block are used to perform texture mapping operations. 
 
 
     
     
       10. The processor as in  claim 9  further comprising:
 a second circuit to generate the decompressed and filtered version of the texture block by performing a dot product of data from the decompression-filtering matrix and the matrix of approximation coefficients. 
 
     
     
       11. The processor as in  claim 10  wherein the second circuit is further configured to:
 reshape the decompression-filtering matrix; and 
 perform the product of the reshaped decompression-filtering matrix and the matrix of approximation coefficients. 
 
     
     
       12. The processor as in  claim 9  wherein the first circuit is further configured to:
 determine a distance matrix [DM] using the distance between each of the plurality of texture block texels and each of the plurality of approximation points; and 
 determine the decompression matrix of member-wise RBF values over the distance matrix [DM] using a specified type of RBF and specified shape parameter e p  value. 
 
     
     
       13. The processor as in  claim 12  wherein the specified type of RBF is selected from a group consisting of Gaussian and Multiquadratic radial basis functions. 
     
     
       14. The processor as in  claim 9  wherein using the decompression matrix to generate the decompression-filtering matrix comprises performing linear combinations on sets of elements from the decompression matrix to arrive at new sets of elements for the decompression-filtering matrix based on the filter function. 
     
     
       15. The processor as in  claim 14  wherein the linear combinations comprise taking an average value of multiple elements from the decompression matrix to determine each element in the decompression-filtering matrix. 
     
     
       16. The processor as in  claim 14  wherein the operations cause the number of elements in the decompression-filtering matrix to be less than, greater than or equal to the number of elements in the decompression matrix.